Evaluación De Los Resultados

[4] Phoenix operations are described in detail by Arvidson

et al. [2009]. In general, soil material was scooped up and imaged by the Surface Stereo Imager (SSI) and Robotic Arm Camera (RAC) [Keller et al., 2008]. The RAC images were most often taken in color mode, i.e., as three successive images with red, green, and blue light emitting diodes (LEDs) switched on, and a fourth one with all LEDs switched off. When imaging the scoop divot (depression cut out near the front of the blade of the scoop at the end of the Robotic Arm (RA) [see Arvidson et al., 2009, Figure 2], high resolution mode (22mm pixel−1) was used. After image documentation, soil samples were transferred to various substrates (micro- bucket, micromachined silicon grid, weakly, or strongly magnetic and sticky silicone) [Hecht et al., 2008; Leer et al., 2008] for further imaging by the Optical Microscope (OM). The OM images (4mm pixel−1) include color information, as they were taken as three successive images with two red, two green, and two blue LEDs switched on. A fourth image with all LEDs switched off was not needed, as the OM is in a light‐ tight box [Hecht et al., 2008]. Unlike the RAC, the OM is a fixed focus camera, and the sample is brought into focus position by the Sample Wheel Translational Stage (SWTS). The in‐focus imaged area is 2 mm × 1 mm. Most of the results on the optical properties of Phoenix soil particles reported in the present paper were obtained from OM images.

[5] The study of particle sizes is based on both OM and

RAC images to facilitate comparisons to size distributions obtained for Gusev soils. Size analysis of Phoenix soil and dust particles down to the submicron scale was done by inclusion of Atomic Force Microscope (AFM) data [Pike et al., 2009; W. T. Pike et al., The particle size distribution of the Martian soil at the Phoenix landing site, manuscript in preparation, 2010]. Images used in the present paper are specified by the mission identifier, an (optional) instrument identifier, the sol number, and the last four digits of the spacecraft clock time. The following instruments (besides the ones onboard Phoenix) have provided data that are used in the present paper: The Microscopic Imager (MI) and the Navi- gation Camera (Navcam) onboard the Mars Exploration Rovers (MER‐A, MER‐B), the Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) onboard the Mars Reconnaissance Orbiter (MRO), and the Wide Field Plane- tary Camera‐2 (WFPC‐2) onboard the Hubble Space Tele- scope (HST).

[6] The spacecraft clock time appears both in the file name

and in the header of the data as archived in the Planetary Data System (PDS; cf. http://pds.jpl.nasa.gov/) and designates the

elapsed since 6 January 2000, 0000:00 UTC. Virtually any image (or RGB image triplet) is uniquely specified by the combination of mission name, sol number, and the last four digits of the spacecraft clock time. For RGB color composites of PHX‐OM images these four digits will refer to the red image.

[7] The calibration pipeline for OM images applies to the

12 bit raw images [Hecht et al., 2008] and involves the fol- lowing steps: (1) bad and hot pixel removal, (2) bias sub- traction, (3) dark current subtraction, (4) conversion to units [DN/s], (5) flat fielding, and (6) conversion to reflectance R*. Steps 1–5 are standard calibration steps [see, e.g., Bell et al., 2006]. Step 5 involves division of the image by a flat field correction array that has a mean value equal to unity. The images that are referred to as RAD calibrated images (in units [DN/s]) have been processed as described in steps 1–5. Step 6 involves the division of the output image from step 5 by a similarly illuminated and calibrated image of the sol 4 white target [Hecht et al., 2008], and a correction for changes in radiant output of the LEDs over the course of the mission. Prior to this division the white target images are flattened by the means of a multiple‐order polynomial fit function. This procedure removes granularity and surface roughness, which are inherent properties of the white target.

[8] Changes of the LEDs’ radiant output as a function of

mission time were inferred from the radiant power reflected by typical dusty regions in the RAD calibrated images. First such regions were selected manually in a large number of images that were acquired throughout the mission. Then the average values of these regions were plotted as a function of sol number and fitted by a straight line.

[9] Obviously this procedure is based on the assumption

that typical dust can be easily recognized in all images and has stable optical properties as a function of time and dig location. The procedure was cross checked against repeated images of the white target (sols 4, 57, 111, 117, 120, 121, and 137) that became increasingly contaminated by dust. It was inferred that the output of the blue, green and red LEDs changed by (+4.1 ± 1.0)%, (+2.0 ± 1.0)%, and (−8.7 ± 2.0)%, respec- tively, over the course of 100 sols. Thus the output of the blue and green LEDs increased very slightly over the course of the mission, while that of the red LEDs decreased by about 10%, which is qualitatively in agreement with premission ex- pectations for the LED types used [Reynolds et al., 2008]. The above numbers were used in calibration step 6.

[10] Strictly speaking we do not know if the signal changes

(as inferred from the RAD images) were caused by changes of the LED output and/or by changes of the CCD sensitivity. These changes could in turn be caused by the number of operational CCD/LED hours or by the temperature of any of these devices. We were unable to distinguish between these different potential causes. Fortunately the above corrections remain valid, whatever may be the causes of these signal changes.

[11] The lighting and viewing geometry for the OM is

largely bidirectional. Each point on a given substrate is generally illuminated by two LEDs of a given color (belonging to clusters 1 and 2; see Figure 1). Given the divergence of the incident light beams (∼10°) the local inci-

GOETZ ET AL.: PROPERTIES OF PHX SOIL PARTICLES E00E22 E00E22

LEDs) and 37.2° (green LEDs). The illumination becomes slightly conical, as the front windows of the LEDs were dif- fused preflight using a fine grinding compound in order to achieve even illumination of the substrates. The reflected light is captured and detected within an approximately 20° wide cone that is centered around the surface normal to the substrate so that emission angles are clustered around an average value of 0° [Hecht et al., 2008]. As mentioned above the incident light arrives from two different directions. However, each illuminating LED should contribute equally to the total reflectance, assuming equal radiant output from both LEDs and assuming random orientation of the light scattering particles on the substrate. Therefore, the lighting‐ viewing geometry may be termed bidirectional rather than tridirectional. Since the images are ratioed to white target images that have been acquired at the very same geometry, the inferred reflectance is R* to the extent that the white target

matches a perfectly diffuse (Lambertian) surface [Reid et al., 1999; Bell et al., 2008].

3.

Analysis

[12] This section describes optical properties (section 3.1),

morphology and surface textures (section 3.2), magnetic properties (section 3.3), and size analysis (section 3.4) of Phoenix soil particles. The optical properties will provide the basis for a classification scheme of the larger (silt to sand sized) grains. In section 3.5 the data presented on Phoenix soil particles will be put into context by comparison to other data sets.

3.1. Optical Properties of Soil Particles

[13] A total of eight soil samples were delivered to the OM

during the mission. Detailed descriptions of sampling loca- tions are provided by Arvidson et al. [2009]. The samples will be discussed in order of delivery to the OM. In some cases, images acquired long after the delivery sol have better qual- ity and were therefore used in the present study (Table 1), although they might be affected by cross contamination among different substrate sets.

3.1.1. Spectral Reflectance

[14] This section describes the optical properties of the

eight strong magnet samples specified above. Figure 2a shows images of the soil samples selected for further anal- ysis. All samples were accumulated on the strongly magnetic substrate (hereafter referred to as the“strong magnet”). In each image (Figure 2b) a region of interest (ROI) containing a thick, in‐focus layer of soil is manually defined. This ROI is termed“all.” Poorly covered or out‐of‐focus regions are discarded from that ROI. The soil samples that were accu- mulated on the strong magnets show in general the largest grain diversity, so they were used for detailed analysis.

[15] Phoenix soil particles are bimodal both in terms of size

and albedo. This very simple statement can be confirmed by visual inspection of Figure 2a. The smallest particles, often unresolved, are predominantly red, while larger grains (typ- ically in the size range 20–100 mm) are dark (mostly brownish or black), and in some cases almost transparent. Most of the larger grains have a significantly flatter (less red) reflectance spectrum than the unresolved fines. According to Wentworth [1922] clay‐sized and silt‐sized particles have maximum diameters of 3.9mm (1/256 mm) and 62.5 mm (1/16 mm), respectively. Grains larger than 62.5 mm are termed sand‐ sized particles. Thus the grains encountered in Phoenix soils

Table 1. Overview of Phoenix Soil Samplesa

Set Sample

Delivery

Sol Images Studied Source Trench and Sample Type

2 Mama Bear 17 R, sol 31, 5549; G, sol 31, 5578; B, sol 31, 5608 Dodo Goldilocks trench, surface sample 1 Rosy Red 26 R, sol 33, 4125; G, sol 33, 4155; B, sol 33, 4184 Rosy Red trench, surface sample 10 Sorceress 38 R, sol 44, 1565; G, sol 44, 1595; B, sol 44, 1625 Snow White trench, scraped pile above ice 8 Mother Goose 67 R, sol 122, 7924; G, sol 122, 7954; B, sol 122, 7984 trench and type of sample unknown 7 Wicked Witch 75 R, sol 122, 0014; G, sol 122, 0044; B, sol 122, 0073 Snow White trench, scraped pile above ice 6 Golden Key 99 R, sol 103, 6251; G, sol 103, 6281; B, sol 103, 6311 Dodo Goldilocks trench, lag deposit scraped pile

above ice

5 Golden Goose 110 R, sol 112, 1338; G, sol 112, 1368; B, sol 112, 1399 Stone Soup trench, subsurface sample

4 Galloping Hessian 128 R, sol 132, 9135; G, sol 132, 9165; B, sol 132, 9195 surface sample below the rock Headless that was flipped over into the Neverland trench on sol 117

Figure 1. Lighting geometry for OM images. Each LED cluster (labeled as 1, 2, or 3) contains three VIS LEDs and one UV LED. The VIS LEDs have been colorized in order to specify the relative position of each LED type. The LEDs are located on two concentric circles around the optical axis with radii of 14.4 mm and 19.1 mm and make an angle of 29.8° and 37.2° with the optical axis, respectively. The dis- tance from the end of the LED to the target is about 25 mm (red and blue LEDs) and 28 mm (green LEDs). Refer also to Hecht et al. [2008]. During the mission the large major- ity of all images were acquired by using LED clusters 1 and 2. LED cluster 3 was rarely used due to undesired scattering effects on the AFM cantilevers.

GOETZ ET AL.: PROPERTIES OF PHX SOIL PARTICLES E00E22 E00E22

are partly silt and partly sand sized. However, for the sake of simplicity we will refer to all of these grains as“sand grains,” although an appreciable fraction of them are silt sized.

[16] Figure 2c shows different ROIs that refer to different

particle types in these soils. These ROIs are largely contained by the above defined ROI“all” and were selected in different ways: The reddish fines (bright red soil or airborne dust) were selected by requiring a reflectance R* < 0.1 in the blue and R* > 0.2 in the red channel, respectively. Among the larger grains two populations could be distinguished by visual inspection in all soil images and these were hand selected: brownish particles with a wide spectral range and black particles.

[17] Figure 3 plots the optical characteristics (in terms of

R* reflectance) of each ROI or type of particles as defined in Figure 2. The reflectance of the ROI“all” (Figure 3a) and that

cedure corrects for a linear overall change in LED radiant power (or camera sensitivity), but does not erase potential fine differences between different soil samples (or substrate sets).

[18] For each particle type (Figures 3b–3d), both the sim-

ple pixel averages (crosses) and the Gaussian fit parameters (open circles) are presented. The center reflectance as inferred from the Gaussian fit is often lower than the corresponding ROI average. This is particularly the case for the red reflec- tance of the fine reddish material (Figure 3b), which in turn is related to the way this material was defined: The requirement was a blue and red reflectance below 0.1 and above 0.2, respectively. This condition is somewhat arbitrary, but allows the selection of pixels dominated by a type of material that is a well‐known and well‐characterized alteration end‐member on the surface of Mars. For many Phoenix soils, most pixels Figure 2. (a) Approximate true color images of Phoenix soil samples on the strongly magnetic substrate.

The images are specified by the sol number, the identifier of the red image (four last digits of the spacecraft clock time), and the substrate set. (b) Region of interest (termed“all”) that is used for the study of the optical properties. The discarded regions (uncovered substrate, out‐of‐focus regions) are blacked out. (c) Further regions of interest that are essentially subregions of Figure 2b. The dusty regions (ROI termed“red fines”) are shown as they appear in the color images. The nonobstructed parts of black and brown grains are repre- sented as uniform white and green regions, respectively. Thus, the individual blotches do not match the out- line of the corresponding particle. The precise definition of these ROIs is given in the text.

GOETZ ET AL.: PROPERTIES OF PHX SOIL PARTICLES E00E22 E00E22

what larger than 0.2, while the Gaussian fit takes the shape of the reflectance histogram into account and provides a center reflectance of about 0.2. Pixels with a reflectance below 0.2 have not been selected, because they do not meet the above conditions and would have likely contained some mixed pixels, such as pixels of brownish sand particles that are

were excellent in most cases and therefore the use of Gaussian fit parameters is favored over simple ROI averages.

[19] The spectral properties of the brownish grains

(Figure 3c) appear to vary between different Phoenix soil samples, and additionally have a large relative uncertainty computed as the standard deviation of the associated ROI. Both facts attest to the large diversity of this type of grains as compared to the black ones (Figure 3d). In the remainder of this paper only a minimum number of categories shall be defined for Phoenix soil particles, and no objective criteria could be found for further subdivision of the brown sand grains.

[20] Figure 4 shows scatter plots of the reflectance and the

spectral slope for all soil samples presented in Figure 2. Pixels in the ROI“all” are shown as small black dots. Pixels covered by the other ROIs shown in Figure 2 have been plotted as black crosses, blue crosses and red dots for the black grains, brownish grains and reddish fines, respectively. Note that pixels belonging to the ROI“all” (Figure 2b), but not to any of the remaining ROIs (Figure 2c), are actually“bad pixels” in the sense of badly illuminated soil patches on the substrate. The number of such bad pixels generally increases with the height and relief of soil material on the substrate. The plots shown in Figures 4g and 4h have a comparatively large amount of such bad pixels. In one case (Figure 4f, or set 6) the region representing the brownish particles overlaps strongly with the one representing the reddish fines. This feature indicates strong contamination of sand particles by reddish fines. It is remarkable that precisely that sample was a sub- limation lag (see Table 1), where finely divided and strongly coloring reddish material was amply available after the freeze dry process.

[21] In summary we have identified and characterized three

different types of soil particles: Reddish (unresolved) fines, dark (almost black) grains, and brownish grains. Hereafter, we will adopt the nomenclature“red fines,” “black sand” and “brown sand,” respectively. These are simple generic terms that are based on a qualitative color analysis. Virtually all particles in the Phoenix (and more generally Martian) soils are characterized by a ferric absorption edge that imparts a reddish color. Obviously the average of all these spectra would be reddish as well. The above naming convention characterizes in a qualitative way the deviation from that average spectrum. The ROIs corresponding to each particle type are named accordingly: “red,” “black” and “brown” designate the ROIs that are associated with red fines, black sand, and brown sand, respectively.

3.1.2. Whitish Particles

[22] In addition to red fines and black and brown sand‐

sized particles, we can distinguish by visual inspection of the images a fourth particle type: flakes of varying extension composed of unresolved whitish particles. A systematic study of brightness histograms also provides evidence for the presence of such particles. The major question to be addressed is the following: How can“whitish particles” be defined in a rigorous way?

[23] The approach we used is based on plotting the number

of pixels that exceed a given reflectance (R*) versus that reflectance. More specifically, this plot (from hereon referred to as threshold plot, TP) is drawn for pixels in the red image. Figure 3. Optical characteristics of Phoenix soil particles.

(a) ROI“all.” (b) ROI “red.” (c) ROI “brown.” (d) ROI “black” (see Figure 2). Each graph displays the R* reflec- tance (red on top, blue on bottom, green between these two) sorted according to substrate set number. In each graph both the ROI averages (crosses) and the Gaussian fits (open cir- cles) are shown. In some cases a meaningful Gaussian fit was not possible due to a low number of data points. The ROI averages are connected by a dotted line in order to guide the eye. Brown sand particles are spectrally diverse and therefore have a strongly varying reflectance with large relative uncertainty. The set number specifies the images that were used for these plots (Table 1).

GOETZ ET AL.: PROPERTIES OF PHX SOIL PARTICLES E00E22 E00E22

plot, TP1, is drawn for all pixels within the ROI“all.” (2) A second one, TP2, is drawn for a subset of pixels within the ROI“all.” All pixels within that subset are required to meet

to 100%. Then TP1 characterizes the actual brightness of all pixels in the given ROI (ROI“all”), while TP2 represents the curve one would expect, if only classical reddish soil or dust Figure 4. (a–d) Scatterplots of the optical properties of Phoenix soil samples. Each row corresponds to the

strong magnet of a given set of OM substrates with the following plot types (from left to right): Blue reflec- tance versus red reflectance, green reflectance versus red reflectance, and green‐red slope of the reflectance versus blue‐green slope. The axis limits are the same for each plot type in order to facilitate comparison between different soil samples. In each graph“all” pixels (Figure 2b) are shown as black dots, whereas pixels covered by the ROIs shown in Figure 2c have been plotted as black crosses, blue crosses, and red dots for the black grains, brownish grains, and reddish fines, respectively. The set number specifies the images that were used for these plots (Table 1). (e–h) Scatterplots of the optical properties of Phoenix soil samples (continued). Each row corresponds to the strong magnet of a given set of OM substrates.

GOETZ ET AL.: PROPERTIES OF PHX SOIL PARTICLES E00E22 E00E22

[24] In Figure 5 TP1 (red, solid line) and TP2 (black,

dashed line) are drawn for the red reflectance of each soil sample and the intersection (Rt*) is determined. The